Search and Discovery¶
Glassy searches across every type you've saved — notes, bookmarks, captures, documents, voice transcripts, and Obsidian vault files. There are two search surfaces: the lightweight universal search overlay and the deep Knowledge Base workspace.
Universal search (Cmd/Ctrl + K)¶
Press Cmd+K (macOS) or Ctrl+K (Windows/Linux) from anywhere in the app. The overlay opens, focused on the input.
What it searches¶
- Note titles, bodies, and frontmatter
- Bookmark URLs, titles, descriptions, and tags
- Capture titles, source URLs, and bodies
- Document text content (PDFs are OCR'd on upload)
- Voice memo transcripts
- Tag names and collection names
Operators¶
| Operator | Example | What it does |
|---|---|---|
tag:<name> |
tag:research |
Filter to items tagged research |
type:<type> |
type:note |
Filter to a content type (note, bookmark, capture, voice, document) |
from:<account> |
from:work |
Limit to a specific account |
after:<date> |
after:2026-06-01 |
Items created on or after the date |
before:<date> |
before:2026-07-01 |
Items created on or before the date |
"exact phrase" |
"second brain" |
Match the exact phrase, not its synonyms |
-<word> |
meeting -standup |
Exclude items containing the word |
Operators combine: tag:research after:2026-06-01 "second brain" is a valid query.
Keyboard navigation¶
| Action | Mac | Windows/Linux |
|---|---|---|
| Open overlay | Cmd + K |
Ctrl + K |
| Next result | ↓ |
↓ |
| Previous result | ↑ |
↑ |
| Open result | Enter |
Enter |
| Filter by type | Tab then arrow keys |
Tab then arrow keys |
| Close | Esc |
Esc |
Knowledge Base workspace (#/kb)¶
The Knowledge Base is the deep search surface — it uses hybrid search (BM25 full-text + vector semantic) with Reciprocal Rank Fusion. It's the same engine the MCP server exposes to your AI tools.
When to use KB instead of universal search¶
- You remember the meaning but not the exact words → use KB. Semantic search finds related concepts even when your query shares no vocabulary with the result.
- You want to search across sub-accounts in one query → use KB. The universal overlay scopes to the active account by default.
- You want relevance scores and source-type filters → use KB. The overlay just shows the top 10 results.
- You want to query your Obsidian vault from inside Glassy → use KB. The universal overlay doesn't search vault files.
KB filters¶
| Filter | Effect |
|---|---|
| Source type | bookmark · note · document · vault_file · voice_transcript |
| Account | limit to one account (default: active) |
| Date range | last 7 days · 30 days · 90 days · custom |
| Sort | relevance · created · updated |
Smart filters¶
Save any KB query as a Smart filter. The filter re-runs automatically when new content lands, so it surfaces as a sidebar entry under Organize → Smart filters. Useful for: "things tagged reading-list I haven't opened in 30 days", "captures from youtube.com this week", "vault files mentioning attention".
Related Notes (on every note)¶
Every note has a Related Notes panel showing the 8 most semantically similar notes in your workspace. It updates as you edit — Related Notes use the same vector embeddings as the Knowledge Base, recomputed on save.
Search across sub-accounts¶
Universal search by default scopes to the active account. To search everywhere you have access, open the KB workspace and pick All accounts in the account filter.
Performance notes¶
- The first KB query after a long idle period may take a beat while the embedding cache warms.
- Vault file search requires the Obsidian Local REST API plugin to be reachable. On the cloud, vault search is unavailable — vault files are imported as notes during ZIP import instead.
- Voice transcript search is full-text; it does not understand audio timestamps. Use the voice memo's inline player to scrub.
Learn more¶
- Second Brain / MCP — the same search engine, exposed to your AI tools
- MCP Server setup — connect Claude, Cursor, Windsurf, OpenCode, and others
- Obsidian Vault Integration — what KB can search in your vault
- Search and the Knowledge Base lesson on Glassy Learn